Develop Reference

MySql Split Query Group Percentage

I am trying to do the following but I cannot manage to get it right yet :(.
I have these tables:
table1 -> tb1_id, tb1_name
Sample Data:
--------------
1 group1
2 group2
3 group3
4 group4
5 group5
table2 -> tb2_id, tb2_sector, tb2_tb3_id
Sample Data:
--------------
1 alpha 1
2 beta 2
3 gamma 2
4 delta 2
5 epsilon 4
table3 -> tb3_id, tb3_mid, tb3_section
Sample Data:
--------------
1 234 alpha,beta,gama,delta
This is the output that I am looking for:
Name Count %
------ ----- -----
group1 1 25%
group2 3 75%
group3 0 0%
group4 0 0%
group5 0 0%
Basically I need a split a column value delimited by a comma (tb3_section in table3) and then find the right group for each value (table2 gives me the group id to link with table1) and then do a total count by group and get the percentage (assuming total is 100%).
This is the query I tried so far:
I searched for split value samples and found one that does the split by creating a numbers table first:
create table numbers (
`n` INT(11) SIGNED
, PRIMARY KEY(`n`)
)
INSERT INTO numbers(n) SELECT #row := #row + 1 FROM
(SELECT 0 UNION ALL SELECT 1 UNION ALL SELECT 2 UNION ALL SELECT 3 UNION ALL SELECT 4 UNION ALL SELECT 5 UNION ALL SELECT 6 UNION ALL SELECT 7 UNION ALL SELECT 8 UNION ALL SELECT 9) t,
(SELECT 0 UNION ALL SELECT 1 UNION ALL SELECT 2 UNION ALL SELECT 3 UNION ALL SELECT 4 UNION ALL SELECT 5 UNION ALL SELECT 6 UNION ALL SELECT 7 UNION ALL SELECT 8 UNION ALL SELECT 9) t2,
(SELECT 0 UNION ALL SELECT 1) t8,
(SELECT #row:=0) ti;
Afterwards, I did this:
select tb3_section, count(1) from (
select
tb3_mid,
substring_index(
substring_index(tb3_section, ',', n),
',',
-1
) as tb3_section from table3
join numbers
on char_length(tb3_section)
- char_length(replace(tb3_section, ',', ''))
>= n - 1
) tb3_section_dashboard
group by 1
This doesn't give me the group count. Just does the split of tb3_section but doesn't give me the correct count and equivalent percentage. Any ideas will be much appreciate it thanks a lot.
LATEST UPDATE
First of all, I would like to thanks #eggyal for pointing me to the right direction and #Shadow for despise knowing that I was not taking the best approach, he came up with a quick fix to my problem. I managed to change the approach and removed the comma delimited values from table3. Instead now I add multiple rows for each new value (and added a constraint to avoid duplicates).
Now table3 looks like:
Sample Data:
--------------
1 234 alpha
2 234 beta
3 234 gama
4 234 delta
5 235 alpha
Here is the query I have taken from #shadow sample:
SELECT t1.tb1_name, COUNT(t3.tb3_section) AS no_per_group,
COUNT(t3.tb3_section) / t4.no_of_groups AS percentage
FROM t1 left
JOIN t2 ON t1.tb1_id=t2.tb2_tb3_id
INNER JOIN t3 ON t2.tb2_sector=t3.tb3_section>0
JOIN (SELECT COUNT(*) AS no_of_groups
FROM t3 INNER JOIN t2 ON t2.tb2_sector=t3.tb3_section>0) t4
GROUP BY t1.tb1_name
Instead of using find_in_set now I use = to match the exact value.
Now I get something like the following but the percentage looks odd and I miss a group that doesn't have a match:
Name no_per_group percentage
----- ------------- ----------
group1 2 0.1053
group3 3 0.1579
group4 3 0.1579
group5 3 0.1579
Although still I need something like:
Name Count %
------ ----- -----
group1 1 25%
group2 3 75%
group3 0 0%
group4 0 0%
group5 0 0%
Notice that if there is no match in a group, I still need to show that group.
Because I have thousands of records which are different from each other, I need to add another condition: where tb3_mid=234 . Likes this, the results are using to tb3_mid.

The best solution would be to redesign your table structure and move the data in the delimited values list to a separate table.
The quick solution is to utilise MySQL's find_in_set() function.
To get the total count of entries in the messages table (table3):
select count(*) as no_of_groups
from t3 inner join t2 on find_in_set(t2.tb2_sector,t3.tb3_section)>0
To get the counts per group, add a join to table1 and group by group name. To calculate the percentage, add the above query as a subquery:
select t1.tb1_name, count(t3.tb3_section) as no_per_group, count(t3.tb3_section) / t4.no_of_groups as percentage
from t1 left join t2 on t1.tb1_id=t2.tb2_tb3_id
inner join t3 on find_in_set(t2.tb2_sector,t3.tb3_section)>0
join (select count(*) as no_of_groups
from t3 inner join t2 on find_in_set(t2.tb2_sector,t3.tb3_section)>0) t4 --no join condition makes a Cartesian join
group by t1.tb1_name

Replacing Id with Actual values in SQL Server

I have the following table structure also I have mention my expected output please help me with query as I don't know much about SQL query
Query :
SELECT * FROM(
SELECT ESIDispensary,ESILocation,test,Category, COUNT(*) AS [Total Count]
FROM
(SELECT category,ESILOCATION,ESIDISPENSARY,TEST
FROM(SELECT id,CompanyId,FName,Code,category,ESILOCATION,ESIDISPENSARY
FROM dbo.[EmployeeDetail] e WHERE e.CompanyId = 1 AND Category in (1,2)) a
LEFT JOIN
(SELECT *
from
(SELECT EmployeeId, CustomeFieldName,FieldValue
FROM dbo.[CustomeFieldDetail] C
JOIN dbo.[EmployeeDetail] e ON e.id = c.employeeid AND e.CompanyId = c.companyid
WHERE e.CompanyId = 1 AND Category IN (1,2)) SRC
PIVOT
(MAX(FieldValue) FOR CustomeFieldName IN([TEST]))
piv)
b ON a.Id = b.EmployeeId
) AS a
GROUP BY ESIDispensary ,ESILocation,test,Category) x
Table generated using above query
ESIDispensary ESILocation test Category Count
12 11 NULL 1 NULL
12 13 30 1 1
14 13 29 2 2
Table 1 : ESI
Id CompanyId FieldName ComboValue
11 1 ESILOCATION mumbai
12 1 ESIDISPENSARY mumbai
13 1 ESILOCATION pune
14 1 ESIDISPENSARY pune
29 1 TEST HDFC
30 1 TEST ICICI
Table 2 : Category
id CategoryName
1 staff
2 manager
Problem is i want to replace IDs with respected values also can i change above query to get expected result
Expected Summary Output :
ESIDispensary ESILocation test staff manager
mumbai mumbai NULL 1 NULL
mumbai pune ICICI 1 1
pune pune HDFC NULL 2

Is this is what you want:
SELECT (SELECT ComboValue FROM ESI WHERE ID = ESIDispensary) AS ESIDispensary,
SELECT ComboValue FROM ESI WHERE ID = ESILocation) AS ESILocation,
SELECT ComboValue FROM ESI WHERE ID = test) AS test,
Category, [Total Count]
FROM
(
SELECT ESIDispensary,ESILocation,test,Category, COUNT(*) AS [Total Count]
FROM
(SELECT category,ESILOCATION,ESIDISPENSARY,TEST
FROM(SELECT id,CompanyId,FName,Code,category,ESILOCATION,ESIDISPENSARY
FROM dbo.[EmployeeDetail] e WHERE e.CompanyId = 1 AND Category in (1,2)) a
LEFT JOIN
(SELECT *
from
(SELECT EmployeeId, CustomeFieldName,FieldValue
FROM dbo.[CustomeFieldDetail] C
JOIN dbo.[EmployeeDetail] e ON e.id = c.employeeid AND e.CompanyId = c.companyid
WHERE e.CompanyId = 1 AND Category IN (1,2)) SRC
PIVOT
(MAX(FieldValue) FOR CustomeFieldName IN([TEST]))
piv)
b ON a.Id = b.EmployeeId
) AS a
GROUP BY ESIDispensary, ESILocation, test, Category) x

Is it possible to define table inside sql statement?

Is there a way to have a custom table inside select statement?
Something like this:
select id from (1, 2, 3, 4) tbl
Assuming tbl is alias and the custom table has one column with 1, 2, 3, 4 rows.
Is this possible without creating a physical temp table?

Yes, you can do it (numbers table) like that in Oracle:
with tbl as (
-- generates table with id column with 1, 2, 3, 4 values
select level id
from dual
connect by level <= 4)
select id
from tbl
you can put the same in different syntax:
select id
from (select level id
from dual
connect by level <= 4) tbl
if you need arbitrary values, say 1, 5, 14, 127 you can achieve it like that:
with tbl as (
select 1 from dual
union all
select 5 from dual
union all
select 14 from dual
union all
select 127 from dual
)
select id
from tbl
Or
select id
from (select 1 from dual
union all
select 5 from dual
union all
select 14 from dual
union all
select 127 from dual) tbl
If you want to have syntax like from (1, 2, 3) you have to declare a type
create or replace type NumberTable is table of number;
...
select *
from table(NumberTable(1, 2, 3, 4)) tbl

create or replace type char_int is table of integer;
/
with numbers as (select column_value
from table(char_int(1,2,5,6,7)))
select numbers.column_value
, object_name
, object_id
from all_objects
join numbers on (substr(object_id,1,1) = numbers.column_value);
;
Output:
<snip>
1 V$HS_AGENT 1999
2 V$HS_SESSION 2001
<snip>

Query to return pairs of related rows, but only when both rows exist

I have the following tables:
*sistema_documentos*
[id], [caminho], [idDocType](FK -> sistema_DocType.id)
*sistema_Indexacao*
[id] ,[idDocumento](FK -> sistema_documentos.id) ,[idIndice](FK ->
sistema_Indexes) ,[valor]
*sistema_DocType*
[id], [tipoNome](FK -> sistema_DocType.id)
*sistema_DocType_Index*
[id],[idName],[mask],[idTipo](FK -> sistema_DocType.id),[tamanho]
From this query:
select distinct a.id, b.idIndice, b.valor from tgpwebged.dbo.sistema_Documentos as a
join tgpwebged.dbo.sistema_Indexacao as b on a.id = b.idDocumento
join tgpwebged.dbo.sistema_DocType as c on a.idDocType = c.id
join tgpwebged.dbo.sistema_DocType_Index as d on c.id = d.docTypeId
where d.docTypeId = 40
and (b.idIndice = 11 AND b.valor = '11111111' OR b.idIndice = 12 AND b.valor = '22222' )
I get the following result
id idIndice valor
13 11 11111111
13 12 22222
14 11 11111111
14 12 22222
16 12 22222
As you can see, I want all ids with idIndice 11 with value 11111111 and 12 with value 22222
Id 16 has id 12 with value 22222 authough it does not have id 11 with value 11111111 so I don´t want it to be shown.
How can I update my query to obtain the result I want. Hope my question is clear. If it is not just ask and I edit my post. Thanks

I would suggest something like this:
WITH TempTable AS
(
select distinct a.id, b.idIndice, b.valor
from tgpwebged.dbo.sistema_Documentos as a
join tgpwebged.dbo.sistema_Indexacao as b on a.id = b.idDocumento
join tgpwebged.dbo.sistema_DocType as c on a.idDocType = c.id
join tgpwebged.dbo.sistema_DocType_Index as d on c.id = d.docTypeId
where d.docTypeId = 40
and (b.idIndice = 11 AND b.valor = '11111111' OR b.idIndice = 12 AND b.valor = '22222' )
)
SELECT *
FROM TempTable t1
WHERE (select count(*)
from TempTable t2
where t1.id = t2.id AND t1.valor != t2.valor) = 1
So... get all the results from your first query where there is at least one result from the table that matches on id, but does not match on valor. (This assumes you could have duplicate rows with the same valor, but you wouldn't want that.)

Try something like this. I took out the tables that didn't have direct bearing on the query, although I named them similarly, and I created a simple schema to replicate the problem. I hope this is clear, and that the connection back to your original query is likewise clear.
CREATE TABLE Documentos (ID INT, document varchar(12))
create table Indexacao (AID INT, indice int, valor varchar(12))
insert Documentos(id, document)
values (1, 'test1'),
(2, 'test2'),
(3, 'test3')
insert Indexacao (aid, indice, valor)
values (1, 11, '11111111'),
(1, 12, '22222'),
(2, 12, '22222')
The important part of the code is the INTERSECT - it returns only rows that are in both sets. In my experience this operator is usually more efficient than anything containing an OR. In the query below, we are getting only those Indexacao rows whose idDocumentos are in the INTERSECT of the two sets of conditions.
SELECT Ind.*
FROM Indexacao Ind
JOIN (
SELECT D.ID
FROM Documentos D
JOIN Indexacao I
ON D.ID = I.AID
WHERE I.Indice = 11 AND I.valor = '11111111'
INTERSECT
SELECT D.ID
FROM Documentos D
JOIN Indexacao I
ON D.ID = I.AID
WHERE I.Indice = 12 AND I.valor = '22222'
)Doc (ID)
ON Doc.ID = Ind.AID
This assumes that you don't have duplicate Indice, Valor rows for a single idDocumento - if you do, you will need to add a DISTINCT.

Selecting multiple row from one row in SQL

I have the following output with me from multiple tables
id b c b e b g
abc 2 123 3 321 7 876
abd 2 456 3 452 7 234
abe 2 0 3 123 7 121
abf 2 NULL 3 535 7 1212
Now I want to insert these values into another table and the insert query for a single command is as follows:
insert into resulttable values (id,b,c), (id,b,e) etc.
For that I need to do a select such that it gives me
id,b,c
id,b,e etc
I dont mind getting rid of b too as it can be selected using c# query.
How can I achieve the same using a single query in sql. Again please note its not a table its an output from different tables
My query should look as follows: from the above I need to do something like
select b.a, b.c
union all
select b.d,b.e from (select a,c,d,e from <set of join>) b
But unfortunately that does not work

INSERT resulttable
SELECT id, b, c
FROM original
UNION
SELECT id, b, e
FROM original
Your example has several columns named 'b' which isn't allowed...

Here, #tmporigin refers to your original query that produces the data in the question. Just replace the table name with a subquery.
insert into resulttable
select
o.id,
case a.n when 1 then b1 when 2 then b2 else b3 end,
case a.n when 1 then c when 2 then e else g end
from #tmporigin o
cross join (select 1n union all select 2 union all select 3) a
The original answer below, using CTE and union all requiring CTE evaluation 3 times
I have the following output with me from multiple tables
So set that query up as a Common Table Expression
;WITH CTE AS (
-- the query that produces that output
)
select id,b1,c from CTE
union all
select id,b2,e from CTE
union all
select id,b3,g from CTE
NOTE - Contrary to popular belief, your CTE while conveniently written once, is run thrice in the above query, once for each of the union all parts.
NOTE ALSO that if you actually name 3 columns "b" (literally), there is no way to identify which b you are referring to in anything that tries to reference the results - in fact SQL Server will not let you use the query in a CTE or subquery.
The following example shows how to perform the above, as well as (if you show the execution plan) revealing that the CTE is run 3 times! (the lines between --- BELOW HERE and --- ABOVE HERE is a mock of the original query that produces the output in the question.
if object_id('tempdb..#eav') is not null drop table #eav
;
create table #eav (id char(3), b int, v int)
insert #eav select 'abc', 2, 123
insert #eav select 'abc', 3, 321
insert #eav select 'abc', 7, 876
insert #eav select 'abd', 2, 456
insert #eav select 'abd', 3, 452
insert #eav select 'abd', 7, 234
insert #eav select 'abe', 2, 0
insert #eav select 'abe', 3, 123
insert #eav select 'abe', 7, 121
insert #eav select 'abf', 3, 535
insert #eav select 'abf', 7, 1212
;with cte as (
---- BELOW HERE
select id.id, b1, b1.v c, b2, b2.v e, b3, b3.v g
from
(select distinct id, 2 as b1, 3 as b2, 7 as b3 from #eav) id
left join #eav b1 on b1.b=id.b1 and b1.id=id.id
left join #eav b2 on b2.b=id.b2 and b2.id=id.id
left join #eav b3 on b3.b=id.b3 and b3.id=id.id
---- ABOVE HERE
)
select b1, c from cte
union all
select b2, e from cte
union all
select b3, g from cte
order by b1
You would be better off storing the data into a temp table before doing the union all select.

Instead of this which does not work as you know
select b.a, b.c
union all
select b.d,b.e from (select a,c,d,e from <set of join>) b
You can do this. Union with repeated sub-select
select b.a, b.c from (select a,c,d,e from <set of join>) b
union all
select b.d, b.e from (select a,c,d,e from <set of join>) b
Or this. Repeated use of cte.
with cte as
(select a,c,d,e from <set of join>)
select b.a, b.c from cte b
union all
select b.d, b.e from cte b
Or use a temporary table variable.
declare #T table (a int, c int, d int, e int)
insert into #T values
select a,c,d,e from <set of join>
select b.a, b.c from #T b
union all
select b.d, b.e from #T b
This code is not tested so there might be any number of typos in there.

I'm not sure if I understood Your problem correctly, but i have been using something like this for some time:
let's say we have a table
ID Val1 Val2
1 A B
2 C D
to obtain a reslut like
ID Val
1 A
1 B
2 C
2 D
You can use a query :
select ID, case when i=1 then Val1 when i=2 then Val2 end as Val
from table
left join ( select 1 as i union all select 2 as i ) table_i on i=i
which will simply join the table with a subquery containing two values and create a cartesian product. In effect, all rows will be doubled (or multiplied by how many values the subquery will have). You can vary the number of values depending on how many varsions of row You'll need. Depending on the value of i, Val will be Val1 or Val2 from original table. If you'll see the execution plan, there will be a warning that the join has no join predicates (because of i=i), but it is ok - we want it.
This makes queries a bit large (in terms of text) because of all the case when, but are quite easy to read if formatted right. I needed it for stupid tables like "BigID, smallID1, smallID2...smallID11" that was spread across many columns I don't know why.
Hope it helps.
Oh, I use a static table with 10000 numbers, so i just use
join tab10k on i<=10
for 10x row.
I apologize for stupid formatting, I'm new here.